CondaEnvException: Pip failed------failed to build tinycudann

[zhanglingyun]

run conda env create --file neuralangelo.yaml and get this How can I solve it

[NickWilde]

I met silimar error yesterday, to me it seems a general error, so you could try doing the following and see if it works:

1. Build Pytorch with cuda before executing conda install.
2. edit neuralangelo.yaml and remove the part with pip.
3. in your .bashrc file export PATH to cuda and LD_LIBRARY_PATH to cuda/lib64
4. After conda finish creating the environment, first activate the environment by conda activate neuralangelo, then run pip install -r ./requirements.txt to continue the installation.

even though it says "your cuda is 11.7 which mismatch the version used to compile Pytorch, it should not be a problem". I still suggest you install a new cuda and use it to compile Pytorch.

Hope this will help

[zhanglingyun]

I met silimar error yesterday, to me it seems a general error, so you could try doing the following and see if it works:

1. Build Pytorch with cuda before executing conda install.
2. edit neuralangelo.yaml and remove the part with pip.
3. in your .bashrc file export PATH to cuda and LD_LIBRARY_PATH to cuda/lib64
4. After conda finish creating the environment, first activate the environment by conda activate neuralangelo, then run pip install -r ./requirements.txt to continue the installation.

even though it says "your cuda is 11.7 which mismatch the version used to compile Pytorch, it should not be a problem". I still suggest you install a new cuda and use it to compile Pytorch.

Hope this will help

Thanks.I give up to use docker then. I 'm not sure whether .bashrc file could set multiple cuda values but I may run many software.Emm,I think docker may be suitable.But thank you again then.